After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 452782 - Image resize sinc / lanczos3 method not working
Image resize sinc / lanczos3 method not working
Status: RESOLVED NOTABUG
Product: GIMP
Classification: Other
Component: General
2.3.x
Other All
: Normal normal
: ---
Assigned To: GIMP Bugs
GIMP Bugs
Depends on:
Blocks:
 
 
Reported: 2007-07-01 10:00 UTC by zigazou
Modified: 2008-01-15 14:12 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
comparison of different scaling methods (205.18 KB, image/png)
2007-07-02 20:01 UTC, Michael Schumacher
Details
Comparison of scaling method on the 2.3.18-devel release (138.00 KB, image/png)
2007-07-03 06:24 UTC, zigazou
Details

Description zigazou 2007-07-01 10:00:23 UTC
Please describe the problem:
There are 4 interpolation types for image resizing : none, linear, bicubic, Sinc/Lanczos3. The last one does not work anymore.




Steps to reproduce:
1. Open a photo (say it's 2048x1536)
2. Resize it down to 104x78 with Interpolation 'Sinc/Lanczos3'


Actual results:
Choosing 'Sinc/Lanczos3' gives similar results to 'none', though not the same.

Expected results:
I think it should look more like the results given by 'Bicubic' than 'none'.

Does this happen every time?
Yes

Other information:
In the beginning of the 2.3 developpement serie, the Lanczos3 interpolation worked like in this release. It worked well in the 2.3.16 release.
Comment 1 Michael Schumacher 2007-07-02 20:01:55 UTC
Created attachment 91053 [details]
comparison of different scaling methods

Not sure if I can reproduce it. See attached image, can you tell which method was used for which one?
Comment 2 zigazou 2007-07-03 06:22:40 UTC
(In reply to comment #1)
> Created an attachment (id=91053) [edit]
> comparison of different scaling methods
> > Not sure if I can reproduce it. See attached image, can you tell which method
> was used for which one?
I'd say :
top left : original
top right : linear or bicubic
bottom left : lanczos3
bottom right : none

I attached a copy of what I have.

Comment 3 zigazou 2007-07-03 06:24:28 UTC
Created attachment 91085 [details]
Comparison of scaling method on the 2.3.18-devel release

The original image shows "Hello world !" phrase at different sizes.
Comment 4 zigazou 2007-07-03 06:37:19 UTC
Additional thought :
I've tried scaling up using Lanczos. It seems to give better results than Bicubic. Would it be that Sinc/Lanczos3 algorithm is dedicated at scaling up and bad at scaling down ?
Comment 5 Michael Schumacher 2007-07-03 08:59:01 UTC
Someone might correct me, but AFAIK the downscaling is still not properly implemented (i.e. the filter kernels are too small) with any interpolation method at all.
Comment 6 Sven Neumann 2007-07-03 11:01:44 UTC
This is being handled in bug #166130 already. Closing the report as NOTABUG because the algorithm does work. It is simply not well suited for decimation (downscaling).